Most productivity advice for founders draws on research that was not conducted on founders, in conditions that do not resemble early-stage company building, to support recommendations that are presented with more certainty than the evidence warrants.
This research digest aims to do the opposite: present what the evidence actually says, flag where it is contested or poorly replicated, and extract the practical implications that hold up even under conservative interpretations.
The Maker’s Schedule: Empirical or Just True?
Paul Graham’s 2009 essay “Maker’s Schedule, Manager’s Schedule” is one of the most-cited pieces of writing in founder productivity discourse. Its core claim — that there are two fundamentally different modes of knowledge work, and that switching between them carries a significant cost — is compelling and rings true to most founders who read it.
It is worth noting that the essay is not a research paper. Graham did not publish data. He described a pattern he had observed across his own experience and the YC founder community.
Does the underlying claim have empirical support? Yes, with nuance.
Research on task-switching costs is substantial and reasonably well-replicated. Work by Rubinstein, Meyer, and Evans (2001) in the Journal of Experimental Psychology found that switching between complex tasks involves two distinct processes — goal shifting (resetting which task you are doing) and rule activation (loading the rules for the new task) — and that both take time, even when the tasks are familiar. The authors estimated that task-switching can consume up to 40% of productive time in scenarios with frequent switching.
More recent work by Sophie Leroy on “attention residue” (2009) found that switching away from an unfinished task leaves behind cognitive preoccupation that impairs performance on the new task. The effect persists even after the physical switch — you can be in a meeting while still partially thinking about the code problem you left unfinished.
These findings support the practical substance of Graham’s claim without requiring the poetic framing. The cost of a meeting inserted into a maker’s afternoon is not just the meeting’s duration. It is the attention residue from the interrupted work and the ramp-up time when you return.
Practical implication: Protect maker blocks from interruption, and — equally important — finish natural stopping points before switching. Leaving a coding problem half-solved to attend a meeting is worse for performance than either completing the problem first or scheduling the meeting before you start.
Cognitive Load and Decision Quality
Founders make dozens of decisions per day, ranging from trivial (which inbox item to address first) to consequential (whether to extend a sales cycle or walk away from a prospect). The quality of those decisions is not uniform across the day.
Daniel Kahneman’s framework from Thinking, Fast and Slow (2011) distinguishes System 1 thinking (fast, intuitive, automatic) from System 2 thinking (slow, deliberate, effortful). High-stakes decisions require System 2 processing. System 2 is subject to depletion.
The ego depletion literature — primarily associated with Roy Baumeister — has had well-documented replication problems. Multiple large-scale attempts to reproduce core ego depletion findings have failed or produced substantially weaker effects than the original studies. The simple “glucose-powered willpower reservoir” model is probably wrong.
What does hold up more robustly is the evidence for cognitive fatigue under sustained mental effort: performance on tasks requiring sustained attention and executive function declines with extended duration (van der Linden, 2011, among others). This is distinct from the specific ego depletion mechanism but produces similar practical implications.
The evidence for time-of-day effects on decision quality is also mixed. The morning advantage for complex cognitive work is real in many studies but not universal — individual chronotype (whether you are naturally a morning or evening person) moderates the effect substantially (Christoph Randler’s work on chronotypes is relevant here, though its direct application to knowledge work quality requires caution).
Practical implication: Schedule your most consequential decisions when you are freshest — which for most people is the first two to three hours after waking. Batch low-stakes decisions and administrative choices into blocks rather than distributing them throughout the day. The mechanism for why this works may not be exactly what the ego depletion literature claimed, but the practical benefit is consistent with multiple lines of evidence.
The Interruption Cost: How Long Does Recovery Take?
Gloria Mark and colleagues at UC Irvine have published extensively on workplace interruptions. Their 2008 work found that after an interruption, workers took an average of 23 minutes to return to the original task at the same level of engagement. A subsequent 2020 study found that the average knowledge worker switches tasks every three minutes when working at a computer.
These numbers are often cited with more precision than they deserve — the 23-minute figure came from a specific organizational context and is not a universal constant. But the directional finding is robust: interruptions are costly, and the cost is much higher than the interruption’s duration alone.
For founders, the implication is structural rather than behavioral. Telling yourself “I will resist interruptions” is a behavioral approach and it works until it does not — stress, urgency, and social obligation erode individual resistance over time. The structural approach is to make interruption physically harder: a closed door, a “do not disturb” status, an explicit communication to team members about availability windows. Mark’s research also found that self-interruptions (checking email voluntarily, opening Slack) are as costly as external interruptions — and more common.
Practical implication: Treat the maker block as a structural intervention, not a personal discipline exercise. The goal is to make interruption structurally difficult, not to resist it through willpower each time.
The Role of Planning Itself in Cognitive Performance
There is a counterintuitive line of research on the relationship between planning and cognitive performance: having a clear plan for upcoming work reduces the attention residue associated with incomplete tasks.
Baumeister and Masicampo (2011) found that participants experiencing intrusive thoughts about unfinished tasks showed improved attention after forming a specific plan to complete them — even without actually completing the task. The mere act of planning appeared to satisfy the “Zeigarnik effect” (the cognitive tendency to remain preoccupied with unfinished work).
This has a direct implication for founders: the 10-15 minutes spent on a structured daily plan is not time taken away from productive work. It may improve the cognitive quality of the work that follows by reducing the background preoccupation with what has not yet been addressed.
The quality of the plan matters, though. A vague intention (“I should work on the authentication module sometime”) does not produce the same cognitive relief as an implementation intention (“I will work on the authentication module from 9am to 12pm in the conference room”). Gollwitzer’s research on implementation intentions (1999) found that specific when-where-how planning significantly improves follow-through and may also reduce the intrusive cognition associated with incompleteness.
Practical implication: A specific daily plan — not just a task list but a schedule with times and contexts — reduces cognitive overhead during the workday. The 10 minutes it takes is a genuine net positive, not a trade-off against productive time.
What Y Combinator’s Data Suggests About Founder Time
Y Combinator does not publish peer-reviewed research, but the accumulated guidance across their essays and office hours represents one of the largest informed bodies of observation about early-stage founder behavior.
Sam Altman’s “How to be Successful” (2019) and Paul Graham’s essays on founder time converge on a few consistent themes:
The founders who build the most successful companies tend to spend a disproportionate fraction of their time talking to users and building product — roughly the Build and Sell modes in the Founder Triangle — and a minimal fraction on organizational overhead in the early stages.
Naval Ravikant’s writing on leverage and time is consistent with this: the founders most capable of building exceptional outcomes tend to be ruthless about protecting the specific type of work where they have unique leverage, and indifferent to the operational overhead that can be delegated, automated, or eliminated.
These are observations and maxims, not controlled studies. The survivorship bias inherent in analyzing successful YC companies — we see what successful founders did, not the full distribution of founder behaviors — means we should interpret these patterns cautiously.
Practical implication: The directional guidance from experienced practitioners who have observed many founders is consistent with the experimental evidence on maker work, interruptions, and cognitive load. Build and Sell work performed in concentrated blocks, protected from interruption, appears to drive the outcomes that matter. The specific numbers (what percentage of time, exactly how long the blocks should be) are less well-established than the directional principle.
Where the Science Is Weakest
Intellectual honesty requires noting where the evidence is thin.
The productivity research base is dominated by laboratory studies, mostly on college-aged participants, performing tasks that are proxy measures of complex cognitive work rather than the work itself. Writing code, doing user interviews, and making architectural decisions are not well-captured by the laboratory tasks used to study cognitive performance.
Studies of actual knowledge workers in naturalistic settings (like Gloria Mark’s workplace research) are more ecologically valid but face their own limitations: the populations tend to be corporate employees, not founders; the tasks observed are often not the deep creative work most relevant to founders; and the Hawthorne effect (behavior change due to being observed) is difficult to rule out.
The specific numbers that circulate in productivity discourse — “it takes 23 minutes to recover from an interruption,” “work in 52-minute blocks,” “the optimal deep work session is 90 minutes” — should be understood as useful heuristics supported by directional evidence, not precise scientific constants.
The honest position: The broad principles hold up. Focused, uninterrupted work on your most important tasks, scheduled when you are cognitively freshest, produces better outcomes than fragmented, reactive work. The precise parameters are individual and should be treated as starting points for self-experimentation, not fixed rules.
The Research Mandate for AI Planning
Given the limitations in the research, the strongest argument for systematic AI-assisted planning is not that it optimally applies proven science. It is that it makes your own data available to you.
No laboratory study captures your specific cognitive patterns, your specific interruption environment, or your specific stage as a founder. Your own weekly planning logs, Triangle audit data, and anchor achievement rates — accumulated over three to six months — are more informative about your specific situation than any general-purpose study.
AI enables this personal data collection with minimal overhead. The weekly close takes five minutes. The monthly trend analysis takes fifteen. The result, over time, is an evidence base about your own productivity that no general research finding can match.
Your action: For the next four weeks, track one number after each maker block: how much uninterrupted time you actually had versus what was planned. At the end of the month, calculate the average gap. That number tells you more about your planning problem than this article does.
Tags: founder productivity research, science of productivity, maker schedule research, cognitive load founders, AI planning evidence
Frequently Asked Questions
-
Is ego depletion real? I've seen conflicting claims.
The ego depletion hypothesis — that willpower draws on a limited mental resource that depletes with use — has had significant replication problems since the mid-2010s. Several large-scale replications of Baumeister's original studies have failed to reproduce the effect. The current scientific consensus is cautious: there may be a fatigue effect associated with sustained self-control, but the original 'glucose-powered willpower reservoir' framing is probably too simple. For practical planning purposes, the useful takeaway remains: schedule your most cognitively demanding work when you are freshest, regardless of the mechanism.
-
What does research say about the optimal length for a deep work session?
Newport's recommendation of 90-minute to four-hour blocks for deep work is consistent with research on ultradian rhythms — roughly 90-minute cycles of higher and lower alertness first described by sleep researcher Nathaniel Kleitman. The practical implication is that 90 minutes is a reasonable minimum for a productive maker block, not because the research proves exactly that number, but because it accounts for the 10–20 minutes of ramp-up time most people need to reach full concentration. Sessions shorter than 60 minutes are often not worth protecting as deep work.
-
Do founders really make worse decisions as the day progresses?
The evidence for decision fatigue in general is mixed, and the most-cited study (Danziger et al., 2011 on parole board decisions) has been challenged methodologically. However, the broader principle — that sustained decision-making over many hours degrades judgment quality — has reasonable support in cognitive science literature on executive function depletion. The practical implication for founders is sound even if the precise mechanism is debated: schedule your most consequential decisions in the first half of your working day.